Dear SPM users,
I received a number of highly informative responses to my question about
temporal lobe susceptibility effects. Since I assume
that they may be of general interest, I list the collected
responses below.
Thanks again.
Peter
Peter Indefrey wrote
Q1)I'm currently considering the pros and cons of PET vs. fMRI for a
language
paradigm with the inferior temporal lobes as the principal region of
interest. While my preference is PET , considering the known
susceptibility problems of fMRI in this region, I'd nonetheless like to
learn what can be done in fMRI to minimize (the effect of) magnetic field
inhomogeneities. Any suggestions?
1)
From: Jose' Ma. Maisog <[log in to unmask]>
Hi Peter, check out this abstract from the recent Human Brain Mapping
conference in San Antonio. They used a post-processing statistical
correction to minimize the effect of susceptibility artifact. Perhaps
one of the authors can offer more advice.
Devlin J, Russell R, Davis M, Price C, Wilson J, Matthews PM, Tyler L,
"Susceptibility and Semantics: Comparing PET and fMRI on a Language
Task," NeuroImage Volume 11, Number 5, May 2000, Part 2 of 2 Parts,
S257.
Q2) Peter Indefrey wrote:
>maybe I should be more specific on this: it was apart from my own
>experience with PET and fMRI on similar paradigms just this talk that
>confirmed my opinion that PET would be the adequate thing to do, since
the
>statistical correction did not seem to fully compensate for the
>susceptibility artifacts. So my question was rather: does everybody agree
>on the conclusions of this paper or are there procedures, post-processing
>or other, such as shimming, choice of slices and angles, that have proven
>effective in minimizing the problem to such an extent that fMRI is at
>least as good as PET in scanning ventral temporal regions.
2)
From: Alejandro Terrazas <[log in to unmask]>
Peter-
I have some experience with field corrections. I am writing up a paper
comparing fieldmap corrected to non-corrected data. Clearly the
raw data is improved but it is still not clear whether there is an
improvement
of the activation maps. For one thing, people use smoothing to "improve"
their images and fieldmap corrections in spiral are like unsmoothing.
Things are different for EPI sequences where you get geometric
distortions. There are correction methods for this as well.
PET vs. fMRI is a tough question. It depends on the temporal dynamics
of what you wish to see. PET is probably more reliable for deeper
structures.
3)
From: Matt Davis <[log in to unmask]>
Hi Peter,
If you read the full paper that has just been published in NeuroImage:
Joseph T. Devlin, Richard P. Russell, Matt H. Davis, Cathy J. Price, James
Wilson,
Helen E. Moss, Paul M. Matthews, and Lorraine K. Tyler (2000)
Susceptibility-Induced Loss of Signal: Comparing PET and fMRI on a Semantic
Task. NeuroImage 11(6): 589^Ö600
you'll see references to three acquisition methods that have been proposed
to alleviate susceptibility-induced problems in the anterior and inferior
portions of the temporal lobe. These are:
1) tailored RF pulses:
Chen, N., and Wyrwicz, A. M. 1999. Removal of intravoxel dephasing
in gradient-echo images using a field-map based RF refocusing
technique. Magn. Reson. Med. 42:807^Ö812.
2) Z shimming:
Constable, R. T. 1995. Functional MRI using gradient echo EPI in
the presence of large static field inhomogeneities. J. Magn. Reson.
Imag. 5: 746^Ö752.
Yang, Q. X., Dardzinski, B. J., Li, S. Z., Eslinger, P. J., and Smith,
M. B. 1997. Multi-gradient echo with susceptibility inhomogeneity
compensation (MGESIC): Demonstration of fMRI in the olfactory
cortex at 3.0 T. Magn. Reson. Med. 37:331^Ö335.
Yang, Q. X., Williams, G. D., Demeure, R. J., Mosher, T. J., and
Smith, M. B. 1998. Removal of local field gradient artifacts in
T-2*-weighted images at high fields by gradient-echo slice excitation
profile imaging. Magn. Reson. Med. 39:402^Ö409.
Constable, R. T., and Spencer, D. D. 1999. Composite image formation
in z-shimmed functional MRI. Magn. Reson. Med. 42: 110^Ö117.
3) Spiral scanning:
Crelier, G. R., Hoge, R. D., Munger, P., and Pike, G. B. 1999. Perfusion-
based functional magnetic resonance imaging with single-shot RARE and
GRASE acquisitions. Magn. Reson. Med. 41:132^Ö136.
As yet we have no practical experience of any of these methods are. There
is a Constable paper in NeuroImage 12(1) showing that z-shimming allows
detection of activations in the hippocampus, but to my knowledge there has
been no equivalent demonstration for activations in the temporal pole. If
the inferior (especially anterior) temporal lobe is a region of interest, I
would agree with you that PET may be a superior imaging modality for your
purposes.
If you hear of any other suggestions other than the ones referred to here I
would be interested to hear of them.
4)
From: Russ Poldrack <[log in to unmask]>
I didn't follow this whole thread, so I'm not sure if anyone has suggested these
steps:
1 - reduce TE to increase the level of bold contrast in the regions where the
dropout occurs - the tradeoff is that you reduce the bold contrast in other
regions. at 1.5T you might try something like 25 ms as a tradeoff
2 - use smaller voxels (something like cubic 3 mm is probably best) - this
definitely reduces the dropout
3 - different orientations may result in different levels of artifact - you
should do test runs with different orientations and see which work best for your
particular area of interest
I've not read the paper mentioned below, but I am skeptical that these artifacts
can be overcome solely using post-processing stats.
5)
From: L.K. Tyler <[log in to unmask]>
In our studies on semantic processing comparing activations in PET and fMRI
(3T) using a similar paradigm and materials, we definitely find more
robust inferior temporal lobe activation in PET than in fMRI. However,
there are procedures you can use to maximise the signal in the temporal
lobes in fMRI, although it's not yet clear how much this will improve
things. For the time being, PET seems to have the advantage if you want to
activate the ventral temporal regions.
6)
From: Joe Devlin <[log in to unmask]>
>From discussions with our physicists, it sounds like there are things
one could do to reduce the macroscopic susceptibility artifacts --
particularly if you
are specifically interested in one region. If the lateral surface of the
ventral temporal region
if the area you are most interested in, you can often use manual shimming
to reduce the
field gradients in that region. The problem is that this typically
distorts the signal from other
areas but if you are doing an ROI analysis, this may not be important.
Obviously, using a
small volume statistical correction for this region will further increase
your sensitivity.
It may be worthwhile piloting your experiment with one or two subjects in
fMRI and
looking at their data before deciding further.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|